Scalability: Set Amazon's Servers on Fire, Not Yours
Don MacAskill, CEO & Chief Geek, SmugMug
his blog: http://blogs.smugmug.com/onethumb
slides posted there, download them
--> great presenation with hard numbers about how to outsource you data storage, processing and hosting <--
SmugMug
192 terrabytes data stored
140 million pictures
doubling every year
$.15 per GB w/ replicas
Why use amazon: not a lot of web-scale expertise on Planet Earth
helps focus on the app, not the muck
Guess: ~ would saved $500k per year
actual
- 64 m to 140 mil photos
- disks $40k to 100k per month
- $922k would have been spent
- $230k spent instead
- $629 in cold, hard savings!
nasty taxes: $295k 'saved' in cash flow! bonus!
- and can resell hard disks, recouping sunk costs
labor mot included, because they had to move 100 TB of data onto
S3 sweet spots
- perfect for startups and small companies
- deal for 'store lots, serve little' business of all sizes
- not so great (yet?) for serving lots if you're a medium or large business. Transfer costs high if you can buy bandwidth
Like SmugFS (file storage) previous:
- everyone building same
-pain to screw in servers
Lessons:
- hot and cold model just right, SmugMug hosts 'hot' data roughly 10% stored locally, s3 holds all of it
- 95% reduction in number of disks bought, no local backup and then only 10% stored local
outages:
3 outages, 15-30 mins, 2 core failures
and one DNS prob. Amazon.com affected
- 2 performance degradations, one our customer noticed, the 2nd they didn't.
- not a big deal, everything fails, expect it.
SLA, Service and Support
Amazon a little weak with customer support
launching ec2 implementation soon
- image processing
- 500k - 1 mil images a day
- 2 terapixles of processing
- ridiculously parallel
Recent Comments